Effect of the Hydrofluoroether Cosolvent Structure ... - ACS Publications

Oct 18, 2017 - Joint Center for Energy Storage Research, 9700 South Cass Avenue, ... Frederick Seitz Materials Research Laboratory, University of Illi...
0 downloads 0 Views 3MB Size
Research Article Cite This: ACS Appl. Mater. Interfaces 2017, 9, 39357-39370

www.acsami.org

Effect of the Hydrofluoroether Cosolvent Structure in AcetonitrileBased Solvate Electrolytes on the Li+ Solvation Structure and Li−S Battery Performance Minjeong Shin,†,‡,# Heng-Liang Wu,†,‡,# Badri Narayanan,†,∥ Kimberly A. See,†,‡ Rajeev S. Assary,†,∥ Lingyang Zhu,‡ Richard T. Haasch,§ Shuo Zhang,†,⊥ Zhengcheng Zhang,†,⊥ Larry A. Curtiss,†,∥ and Andrew A. Gewirth*,†,‡ †

Joint Center for Energy Storage Research, 9700 South Cass Avenue, Argonne, Illinois 60439, United States Department of Chemistry and §Frederick Seitz Materials Research Laboratory, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, United States ∥ Materials Science Division and ⊥Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, United States ‡

S Supporting Information *

ABSTRACT: We evaluate hydrofluoroether (HFE) cosolvents with varying degrees of fluorination in the acetonitrilebased solvate electrolyte to determine the effect of the HFE structure on the electrochemical performance of the Li−S battery. Solvates or sparingly solvating electrolytes are an interesting electrolyte choice for the Li−S battery due to their low polysulfide solubility. The solvate electrolyte with a stoichiometric ratio of LiTFSI salt in acetonitrile, (MeCN)2− LiTFSI, exhibits limited polysulfide solubility due to the high concentration of LiTFSI. We demonstrate that the addition of highly fluorinated HFEs to the solvate yields better capacity retention compared to that of less fluorinated HFE cosolvents. Raman and NMR spectroscopy coupled with ab initio molecular dynamics simulations show that HFEs exhibiting a higher degree of fluorination coordinate to Li+ at the expense of MeCN coordination, resulting in higher free MeCN content in solution. However, the polysulfide solubility remains low, and no crossover of polysulfides from the S cathode to the Li anode is observed. KEYWORDS: lithium−sulfur battery, solvate electrolyte, sparingly solvating electrolyte, hydrofluoroether cosolvent, variable-temperature NMR spectroscopy, X-ray photoelectron spectroscopy



within the S cathode,9−11 fabricating a protecting layer on the Li anode,12 electrolyte additives to modify the reaction pathway and passivate the Li anode,13−15 and tuning the electrolyte composition to reduce the solubility of polysulfides.16,17 Among the electrolyte modification approaches, the use of a high Li salt concentration is effective in minimizing polysulfide dissolution, mitigating the polysulfide shuttle, and achieving stable battery cycling.18−21 The low polysulfide solubility in concentrated electrolytes can be explained by the common ion effect where the equilibrium between Li2Sn and solvated Li+ is shifted to disfavor solvated Li+ as a result of the excess solvated Li+ in solution (Li2Sn ⇌ Li+ + Sn2−).18 In addition to the common ion effect, the unique solution structure of the concentrated electrolyte may also dictate the polysulfide solubility. The Li+ solvation structure of the concentrated

INTRODUCTION With the increasing demand for electric vehicles and portable electronics, Li−S batteries are receiving considerable attention as promising next-generation batteries due to their high energy density and the low cost of sulfur.1,2 The S cathode delivers a theoretical capacity of 1675 mA h g−1, compared to only 274 mA h g−1 for the LiCoO2 intercalation cathode used in the Li ion battery.3 A major problem hindering Li−S battery commercialization is the complicated reduction/oxidation pathway of the Li−S chemistry. The electrochemical conversion from S8 to Li2S involves a phase change during which soluble lithium polysulfide intermediate species are generated.4−6 High-order lithium polysulfides (Li2Sn, n ≥ 4) readily dissolve into the organic electrolyte, causing severe capacity fading and poor Coulombic efficiency.3 In addition, the migration of intermediate polysulfides to the Li metal anode causes the well-known polysulfide shuttle effect.7,8 Various strategies are employed to address the issue of polysulfide dissolution, including physical confinement of polysulfides © 2017 American Chemical Society

Received: August 3, 2017 Accepted: October 18, 2017 Published: October 18, 2017 39357

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces Table 1. Hydrofluoroether Cosolvents Used for (MeCN)2−LiTFSI Electrolyte Dilution

viscosity and increase the ionic conductivity.21 Besides its role in improving the transport properties, the TTE has a significant effect on the local solvation structure of the electrolyte and the Li−S electrochemistry.25 Previous work shows that adding TTE to the (MeCN)2−LiTFSI electrolyte results in TTE coordination to Li+ at the expense of MeCN coordination, releasing free or uncoordinated MeCN.25 Increased free MeCN facilitates the S8 reduction kinetics at the cathode by local solvation effects, which is likely the cause for the enhanced battery cycling behavior and high S8 utilization efficiency.25 Additionally, HFEs are known to affect the Li−S battery performance by suppressing the polysulfide shuttle effect, mitigating selfdischarge, and affecting solid−electrolyte interphase (SEI) formation, overall improving the cyclability and Coulombic efficiency.16,28−34 Further studies by Balasubramanian and coworkers showed that a moderate rise in Li−S cell operating temperature again redirects the reaction pathway, allowing for minimal overpotential and high sulfur utilization efficiency.27 The fine-tuning of polysulfide solubility by temperature control alters the conversion of S8 to Li2S to follow a quasi-solid-state reaction where intermediate polysulfide species act as internal redox mediators and change the relative rate of competing reactions.27 Due to the intriguing ability of the TTE to affect the local structure of Li+ and consequently the S8 electrochemistry, we set out to understand the origin of these interactions and develop structure−property relationships by modifying the HFE structure. This approach has allowed us to determine if there is a specific molecular-scale interaction we can utilize to design advanced HFE-containing electrolytes. To evaluate interactions that might change the Li−S battery properties, herein we examine four different HFEs with varying degrees of fluorination as cosolvents in the (MeCN)2−LiTFSI electrolyte. The use of different HFEs enables fine-tuning of the Li+− solvent interaction that affects the cycling stability of the Li−S cell. Cycling of the Li−S battery shows that the addition of highly fluorinated HFEs yields higher discharge capacity relative to that of cells with less fluorinated HFE cosolvents. To understand this effect, we evaluate the physicochemical properties, solvation structure, polysulfide solubility, and SEI formation on the Li anode using Raman spectroscopy, NMR spectroscopy, ab initio molecular dynamics simulations, UV−

solution is completely different from that of the dilute solution in which solvent-separated ion pairs dominate the salt speciation.22,23 As the Li salt concentration is increased, solvent-separated ion pairs are forced into contact ion pairs in which the Li coordination sphere is composed of both the salt anion and the solvent molecules.22,23 When the Li salt concentration is increased to a level in which all solvent molecules are coordinated to Li+ to form the “solvate” complex, only a small amount of free solvent is present to solubilize the polysulfide species, significantly decreasing the polysulfide solubility.20 This class of electrolyte is called the solvate or sparingly solvating electrolyte. Watanabe and co-workers evaluated the polysulfide solubility and corresponding electrochemical performance of the Li−S cell in glyme-based solvate electrolytes, where higher salt concentration electrolytes exhibit a longer cycle life and higher Coulombic efficiency.20 Similarly, Nazar and co-workers studied solvate electrolytes for the Li−S battery, based on a high concentration of lithium bis[(trifluoromethyl)sulfonyl]imide (LiTFSI) salt in acetonitrile (MeCN) to form the solvate (MeCN)2−LiTFSI.21 The low polysulfide solubility is achieved by the unique solvation structure of the complex where the Li+ cation is chelated by solvent molecules, thereby decreasing its donating ability toward polysulfides.20,21,24 Despite these favorable effects, the solvate electrolytes exhibit high viscosity and low ionic conductivity compared to the conventional 1 M ethereal electrolytes.20,21 Cosolvents such as hydrofluoroether (HFE) that solubilize the solvate complexes but maintain limited solubility of the lithium polysulfides have been used to lower the viscosity of the electrolytes while maintaining beneficial effects on the electrochemical behavior.20,21 The (MeCN)2−LiTFSI solvate electrolyte has been a special focus of recent studies, due to its intriguing solution structure and unique discharge profile compared to the curve obtained in conventional ether-based electrolytes.21,22,25−27 The Li−S cell with (MeCN)2−LiTFSI electrolyte exhibits a sloping voltage profile with two plateaus that are less clearly defined, suggesting a fundamentally different reaction mechanism.21 However, the Li−S cell with neat solvate experiences severe capacity fading and poor Coulombic efficiency.21 Stable cycling is achieved only when the HFE cosolvent, 1,1,2,2-tetrafluoroethyl 2,2,3,3tetrafluoropropyl ether (TTE), is added to decrease the 39358

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

plating/stripping experiments were performed at room temperature (∼23 °C). Physicochemical Property Measurement. The viscosity of the electrolyte was measured with a VISCOlab 4000 (Cambridge Viscosity by PAC Corporate) at 25.0 ± 0.02 °C, and the temperature was equilibrated by a CF41 Cryo-Compact circulator (JULABO USA Inc.). The ionic conductivity was determined by ac impedance spectroscopy (BioLogic SP-150) in a symmetric cell at room temperature (∼23 °C). The density of the electrolyte was measured in a 1.00 mL volumetric flask. Raman Spectroscopy Measurement. Raman spectroscopy of the electrolytes was measured in 5 mm NMR tubes sealed under argon with an experimental setup reported previously.35,36 The measurement was carried out at room temperature (∼23 °C). A 50 mW 632.8 nm He−Ne laser (Meredith Instruments) was used for sample excitation, and the laser power at the sample was 10 mW. The spectral resolution was ∼3 cm−1, and a good signal to noice ratio (S/N) was achieved from an acquisition time of 240 or 480 s per spectrum. Curve fitting was performed using a Gaussian function to determine the peak area. NMR Spectroscopy Measurement. 7Li NMR measurements were performed on a 600 MHz Varian instrument in 5 mm screw-cap NMR tubes with a 5 mm AutoX DB probe with Z-gradient capability at the University of Illinois at Urbana-Champaign School of Chemical Sciences (UIUC SCS) NMR Laboratory. External standards containing 10 M LiCl in D2O were inserted into each sample in the form of a coaxial sealed capillary. The 90° pulse width was calibrated for each sample at every temperature. The T1 longitudinal relaxation time was measured using the inversion−recovery pulse sequence. The T1 values were calculated by fitting the intensity of the peak as a function of the relaxation delay to an exponential function. The 7Li T1 measurements were performed over a wide temperature range (from −35 to +70 °C) to observe the T1 minimum. Between each temperature point, the samples were allowed to rest for 15 min to achieve thermal equilibrium. 15N NMR measurements were also performed on a 600 MHz Varian instrument in 5 mm screw-cap NMR tubes at the UIUC SCS NMR Laboratory. The 15N NMR spectra were obtained at room temperature (∼23 °C). 15N chemical shifts were referenced to an external standard, NH415NO3, in D2O (at −4 ppm) introduced to each sample via a coaxial sealed capillary. Ab Initio Molecular Dynamics Simulation. We performed ab initio molecular dynamics (AIMD) simulations using a density functional theory (DFT) projector-augmented wave (PAW) formalism within the generalized gradient approximation (GGA) as implemented in the Vienna Ab-initio Simulation Package (VASP).37,38 The bulk electrolytes were simulated using a computational supercell (18 Å × 18 Å × 18 Å) consisting of ∼400 atoms, with periodic boundary conditions along all directions. Five different electrolyte systems were simulated, including the neat (MeCN)2−LiTFSI solvate as well as the (MeCN)2−LiTFSI:HFE diluted with four different HFEs (i.e., TTE, OTE, BTFE, and ETE) at a 1:1 volume ratio. In the simulations, the density of the electrolyte was fixed close to experimentally measured values (see Table S1 in the Supporting Information for the measured density values). We employed the PAW pseudopotentials supplied by VASP, while the exchange correlation was described by the Perdew− Burke−Ernzerhof functional.39 Long-range van der Waals dispersion interactions were treated using the DFT-D2 method of Grimme.40 The plane-wave energy cutoff was set at 520 eV, while the Brillouin zone is sampled at the Γ-point only. For each of the bulk electrolyte systems, the initial configuration was made up of LiTFSI, MeCN, and HFE molecules placed at random locations with arbitrary orientations. Each system was equilibrated at 300 K in the canonical ensemble (NVT)41 for 6 ps with a time step of 0.5 fs. A constant-temperature condition was maintained using a Nose−Hoover thermostat. The solvation structure around the Li+ ions (made up of all molecules that contain at least one atom within 2.7 Å of the Li+ ion) was averaged over the last 1 ps of the AIMD runs. Quantum Chemical Cluster Calculations. The thermodynamic properties of the Li+ solvation structure obtained from AIMD simulations were further investigated using DFT quantum chemical (QC) cluster calculations as implemented in the Gaussian 09 code.42

vis spectroscopy, and X-ray photoelectron spectroscopy (XPS). Raman and NMR spectroscopies show that HFEs with a higher degree of fluorination coordinate to Li+ at the expense of MeCN coordination, producing relatively higher free MeCN content in solution. In addition, the solvate electrolytes diluted with highly fluorinated HFEs exhibit lower polysulfide solubility, resulting in cleaner Li metal anodes with fewer polysulfide byproducts. This work provides design rules for developing an advanced solvate electrolyte which is sparingly solvating with respect to polysulfides while maintaining reactivity at the cathode.



EXPERIMENTAL SECTION

Electrolyte Preparation. Lithium bis[(trifluoromethyl)sulfonyl]imide (LiTFSI) salt was purchased from Sigma-Aldrich and dried at 130 °C under vacuum for 8 h. Anhydrous acetonitrile (MeCN; 99.8%, Sigma-Aldrich), 1,1,2,2-tetrafluoroethyl 2,2,3,3-tetrafluoropropyl ether (TTE; 99%, Synquest Laboratories), 1H,1H,5H-octafluoropentyl 1,1,2,2-tetrafluoroethyl ether (OTE; 97%, Synquest Laboratories), bis(2,2,2-trifluoroethyl) ether (BTFE; 99%, Synquest Laboratories), and ethyl 1,1,2,2-tetrafluoroethyl ether (ETE; 99%, Synquest Laboratories) were dried over activated alumina for 7 days. The chemical structures of all HFEs are shown in Table 1. The solvate electrolyte, (MeCN)2−LiTFSI, was prepared by stirring a stoichiometric ratio of 2 mol of MeCN and 1 mol of LiTFSI overnight in an argon-filled glovebox to yield a clear, colorless, viscous solution. The electrolytes with HFE cosolvents were prepared by diluting the (MeCN)2−LiTFSI electrolyte with the corresponding HFE at volume ratios of 2:1, 1:1, and 1:2 (solvate:HFE). All electrolytes were prepared in an argon-filled glovebox. The water content of the electrolytes measured with a Photovolt Aquatest Karl Fischer Coulometric titrator is less than 10 ppm in all cases. Two-Electrode Swagelok Cell Preparation. S8-infiltrated mesoporous carbon (S@CMK-3) was prepared by melt diffusion of S8 (99.98%, Sigma-Aldrich) in ordered mesoporous carbon, CMK-3 (BET-1000, ACS Material), at 150 °C as described previously.9 The S@CMK-3 composite material was prepared at 50 wt % S8. The cathode slurry consisting of 80 wt % S@CMK-3, 10 wt % carbon black (Super P Li, Timcal Inc.), and 10 wt % poly(vinylidene fluoride) (PVDF; Kynar 2801) binder was mixed with anhydrous N-methyl-2pyrrolidone (NMP; Sigma-Aldrich). The slurry was cast onto an aluminum substrate and dried overnight at 55 °C. The electrodes were then punched into 0.5 in. diameter electrodes to yield cathodes with an S8 loading of 0.24−0.4 mg cm−2. Li−S cells were assembled in a modified Swagelok tube apparatus (nylon, 0.5 in. inner diameter, Chicago Fluid System Technologies). The two-electrode Swagelok cell consisted of a Li metal anode (99.99%, Alfa Aesar), a 2400 Celgard separator, and the S@CMK-3 cathode. The electrolyte volume was controlled at 20 μL, resulting in an S8 mass to electrolyte volume ratio of ∼20 g L−1. Electrochemical Measurements. The Li−S batteries were cycled in the Swagelok cell at a rate of 0.05 C (calculated using the weight of S8) for the first two cycles and 0.1 C for the following cycles using an Arbin battery tester (model BT 2043, Arbin Instruments Corp., United States). To ensure electrode wetting, the cells were allowed to rest at open circuit for at least 12 h before cycling. The Li−S cell cycling was performed at room temperature (∼23 °C). The Coulombic efficiency (CE) was determined by CE = Qdischarge,(n+1)th cycle/Qcharge,nth cycle × 100. The potential window for battery cycling was controlled between 3.0 and 1.5 V (vs Li/Li+). All potentials are referenced to Li/Li+. The cycling stability of the Li metal anodes was evaluated in Li−Li symmetric Swagelok cells fitted with 2400 Celgard separators. The electrolyte volume was 30 μL. Galvanostatic cycling was performed on an Arbin battery tester (model BT 2043, Arbin Instruments Corp.) by applying a constant current of ±0.5 mA cm−2 for 1 h in each half-cycle. The amount of Li plated/stripped was 0.5 mA h cm−2. The Li metal 39359

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

Figure 1. (a) Electrochemical performance of Li−S cells prepared with a Li metal anode, an S@CMK-3 cathode, and a neat (MeCN)2−LiTFSI solvate electrolyte diluted with HFEs at a volume ratio of 2:1 (solvate:HFE). The neat (MeCN)2−LiTFSI solvate electrolyte is also shown for comparison. The cells are cycled between 1.5 and 3.0 V at 0.05 C for the first two cycles and 0.1 C for the following cycles. The Coulombic efficiency (CE) shown here is the discharge efficiency and was determined by CE = Qdischarge/Qcharge × 100. (b) Corresponding charge and discharge curves for cycle 10. The Li−S cell cycling was performed at room temperature. The solvation cluster in the first coordination layer around Li+ in the neat (MeCN)2−LiTFSI is considered first for the structure and energy evaluations. From these species, the thermodynamic feasibility of replacing one MeCN molecule in the solvate structure with different HFEs was computed. To accomplish this, we optimized several geometries containing one TFSI− anion, one MeCN molecule, and one molecule of a particular HFE around Li+, including those sampled from our AIMD trajectories. In each of these cases, the long-range corrected hybrid functional, ωB97x-D,43 with the 6-31+G(d,p) basis set, was used to optimize the geometry and to evaluate the electronic energy (E), enthalpy (H), and Gibbs free energy (G) in the gas phase. For all species, a single-point energy calculation (Esolvent) using the conductor-like polarizable continuum model (CPCM)44 was performed using acetonitrile medium (ε = 36.6) at the ωB97x-D/631+G(d,p) level of theory to approximate the solvation energy (ΔE = Esolvent − Egas). The enthalpy of binding Li salt with solvent molecules in solution (Hsoln) was approximated by taking the sum of the gasphase enthalpy of binding (ΔHgas) and the change in solvation energy upon binding (ΔEsolvent). Polysulfide Solubility Measurement. A solid form of “Li2S8” was prepared by dissolving a stoichiometric ratio of S8 and Li2S in MeCN with stirring at 70 °C overnight. The MeCN solvent was removed by placing the polysulfide solution under an active vacuum for ∼8 h. The solvate/HFE solutions saturated with Li2S8 were prepared by adding 50 mM Li2S8 to solvate/HFE mixtures and stirring at 70 °C for 24 h followed by a rest period at room temperature for at least 72 h prior to the measurements. After the rest period, a precipitate of undissolved solid was observed. Therefore, the solvate/ HFE solutions were assumed to be saturated with lithium polysulfide.20 The supernatant solutions were then interrogated by UV−vis spectroscopy (Cary 5000 UV−vis−NIR spectrophotometer). A baseline correction was performed to avoid any absorbance from the neat solvate/HFE solutions. Surface Characterization. X-ray photoelectron spectroscopy (XPS) was performed by using a Kratos AXIS Ultra spectrometer equipped with an Al Kα (1486.6 eV) X-ray source. All binding energies were referenced to the C 1s peak at 284.8 eV. The XPS spectra were analyzed by subtracting the background using the Shirley method and fitting with a Voigt function (70% Gaussian/30% Lorentzian). Li metal samples were prepared by disassembling the Li−S cell after the 100th cycle while the cell was in the fully charged state, followed by rinsing thoroughly with 1,3-dioxalane (DOL) solvent, and drying under vacuum overnight before characterization. The samples were exposed briefly to air when being transferred into the vacuum chamber for the XPS measurement.

about the ether were evaluated as cosolvents in the (MeCN)2− LiTFSI solvate electrolyte to determine the effect of the HFE structure on the electrochemical performance of the Li−S battery. The structures, molecular weights, and boiling points of each HFE are shown in Table 1, including the previously studied TTE,20,21,25,27 BTFE,30,31 and ETE.33 The HFEs were chosen due to their commercial availability and diverse functionality. For example, the degree of fluorination is highest in the OTE (75%), followed by TTE (66.6%), BTFE (60%), and finally ETE (40%). ETE is the only HFE containing a terminal −CH3 moiety, and BTFE is the only HFE with a terminal −CF3 moiety. BTFE is also the only symmetric HFE. The electrochemical performance of the Li−S cells with HFE-diluted solvate electrolytes was evaluated by galvanostatic cycling experiments. Figure 1 shows the discharge capacity and Coulombic efficiency (CE) as a function of the cycle number along with representative discharge and charge curves using an S8-infiltrated mesoporous carbon cathode (S@CMK-3). The capacity fade as a function of the cycle number is alleviated upon addition of any of the four HFEs relative to that of the HFE-free solvate electrolyte, consistent with prior results using TTE as a cosolvent.21 The degree of capacity retention, however, differs between each HFE, with the TTE and OTE yielding higher discharge capacities at 100 cycles relative to those of cells in which BTFE and ETE were added to the solvate electrolyte. At cycle number 100, TTE- and OTEdiluted electrolytes exhibited discharge capacities of 686 and 546 mA h g−1, whereas the BTFE and ETE cosolvents resulted in capacities of only 195 and 148 mA h g−1, respectively. The corresponding Coulombic efficiencies for electrolytes with HFE cosolvents were all ca. 100%, whereas the (MeCN)2−LiTFSI electrolyte itself exhibits poor Coulombic efficiency with larger deviations from 100%. The 2:1 solvate:HFE volume ratio shown in Figure 1 produced better cycling behavior compared to a 1:1 ratio and was thus determined to be the optimal composition. The cycling performance of Li−S cells with electrolytes at a solvate:HFE ratio of 1:1 is shown in Figure S1 (Supporting Information). For the 1:1 system in which the relative amount of HFE is increased compared to that in the 2:1 system, the beneficial effect of TTE and OTE addition is less pronounced and the Coulombic efficiency becomes unstable, exhibiting greater deviations from 100%. Additionally, the capacity fade is more severe and the measured discharge capacities are lower compared to those of the 2:1 systems. Thus, the HFE content



RESULTS AND DISCUSSION Electrochemical Performance of Li−S Cells with the HFE-Diluted (MeCN)2−LiTFSI Solvate Electrolytes. HFE cosolvents with varying degrees of fluorination and symmetry 39360

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

Table 2. Physicochemical Properties of the (MeCN)2−LiTFSI Solvate Electrolyte and Its HFE Diluents at a 2:1 Volume Ratio

a

electrolyte

MeCN:LiTFSI:HFE mole ratio

density (g mL−1), rt

LiTFSI conc (M)

ionic conductivity (mS cm−1), rt

viscosity (cP), 25 °C

neat solvate solvate:TTE (2:1) solvate:OTE (2:1) solvate:BTFE (2:1) solvate:ETE (2:1)

2:1:0 2:1:0.818 2:1:0.606 2:1:0.946 2:1:1.005

1.459 1.465 1.503 1.436 1.376

3.952 2.620 2.635 2.651 2.667

1.35a 1.54 1.16 2.13 1.88

140.2 24.99 37.77 17.34 16.34

This value is taken from ref 21.

Figure 2. Raman spectra of neat (MeCN)2−LiTFSI solvate electrolyte diluted with (a) TTE, (b) OTE, (c) BTFE, and (d) ETE with volume ratios of 2:1, 1:1, and 1:2 (solvate:HFE). The black dashed line shown in (a) represents the curve-fitting results for the neat solvate. The lower wavenumber region shows the TFSI−-related modes a and a′, and the high wavenumber region shows bands associated with MeCN, modes b, b′, c, and c′. The spectra are normalized and overlaid for comparison. The intensity around mode b (2230−2260 cm−1, highlighted with yellow) is multiplied by a factor of 4. (e) HFE addition to the (MeCN)2−LiTFSI solution releases free MeCN, as determined by the peak area ratios of the free MeCN mode to coordinated MeCN (mode b vs mode b′). The Raman spectra of the solvate/TTE solutions have been published previously25 and are reproduced here. The Raman spectra shown here were obtained at room temperature. Adapted from ref 25. Copyright 2016 American Chemical Society.

OTE-diluted electrolytes. Thus, the transport properties of the HFE-containing electrolytes are not the cause for enhanced battery cycling performance between the four HFEs. Solution Structure of the HFE-Diluted MeCN Solvate Electrolytes. Evaluating the local structure of the electrolyte is important to understand the unique behavior of the HFEdiluted solvate electrolyte in the Li−S battery. Previous work has shown that the Li+ in the (MeCN)2−LiTFSI electrolyte is highly coordinated by the TFSI− anion and MeCN solvent molecules to form a unique solvation structure that differs from that of the dilute solutions.25,45 Addition of the HFE to the solvate electrolyte was suggested to be inert toward the Li+ solvation structure;21 however, we have shown that addition of TTE to the (MeCN)2−LiTFSI changes the Li+ coordination structure.25 A fraction of the TTE coordinates to the Li+ at the expense of the MeCN coordination in the solvate complex, thereby releasing uncoordinated, or “free”, MeCN in solution.25 Because the TTE cosolvent has a significant effect on the local solvation structure of the electrolyte solution, here we determine how different HFEs modify the solution structure of the solvate electrolyte using Raman and NMR spectroscopy along with ab initio molecular dynamics simulations. Raman Spectroscopy. To interrogate the Li+ coordination environment in the HFE-added solvate electrolytes, Raman spectra of (MeCN)2−LiTFSI with the addition of HFEs at volume ratios of 2:1, 1:1, and 1:2 were obtained. Parts a−d of Figure 2 show the overlaid Raman spectra in two wavenumber regions of each electrolyte normalized to the largest peak

also has a significant effect on the Coulombic efficiency and the discharge capacity of the battery. Figure 1b shows the discharge and charge profiles for the Li− S cells at cycle 10. The features in the discharge and charge profiles measured in electrolytes with HFE are similar to those measured in the neat (MeCN)2−LiTFSI electrolyte in that two discharge plateaus are observed. The cell with the OTE solvate has a slightly lower first discharge plateau and slightly higher charge plateau compared to other HFE solvates, suggesting higher overpotentials for both reduction and oxidation. The larger cell polarization in OTE is likely due to the relatively high viscosity of solvate:OTE (2:1) electrolyte compared to the other HFEs, which causes sluggish kinetics (Table 2). Origin of the HFE Effect. We next evaluate the origin of the improved cycling behavior of the Li−S cell with the addition of TTE and OTE to the (MeCN)2−LiTFSI solvate electrolyte relative to the addition of the other two HFEs. We consider several possible causes for the improvement, including viscosity/conductivity effects, the solution structure, the polysulfide solubility, and SEI on the Li metal anode. Physicochemical Properties. Table 2 reports the ionic conductivity and viscosity of the (MeCN)2−LiTFSI solvate electrolyte and the solvate diluted with the four HFEs. The BTFE- and ETE-diluted electrolytes exhibit significantly higher ionic conductivities and lower viscosities relative to TTE and OTE-containing electrolytes. The cells cycled with the BTFEand ETE-diluted electrolytes, however, exhibit the most severe capacity fade compared to the cells cycled with the TTE- and 39361

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

formed in solution. We have previously shown that, as TTE is titrated into (MeCN)2−LiTFSI, TTE displaces coordinated MeCN, increasing the relative amount of free MeCN.25 The same trend is observed with other HFEs, including OTE, BTFE, and ETE. As the HFE content is increased, from 33% at a 2:1 ratio and 50% at a 1:1 ratio to 67% at a 1:2 ratio, the peak area of mode b increases relative to that of mode b′, indicating that addition of HFE releases free MeCN into the solution (Figure 2a−d). For quantitative analysis, curve fitting was performed to determine the relative peak area ratios of free MeCN and coordinated MeCN (Figure S3, Supporting Information). The fraction of free MeCN relative to coordinated MeCN versus the HFE content is shown in Figure 2e, where the Raman peak area ratio of free MeCN (mode b) to coordinated MeCN (mode b′) increases linearly as a function of the HFE content. The linear relationship is observed with all HFEs tested, with minimal differences between each HFE. We note, however, that the resolution of the Raman measurement is relatively low, making it difficult to quantitatively determine small changes in the free MeCN content between the HFEs at the same volume ratios. We can, however, observe measurable differences as the volume of HFE is increased, with the free MeCN content increasing as the HFE content increases. Variable-Temperature 7Li NMR Spectroscopy. To directly probe the coordination environment around the Li+ cation, 7Li NMR measurements of the (MeCN)2−LiTFSI and its HFE diluents were performed. Figure 3 shows the 7Li longitudinal

intensity. The spectra without normalization are also shown in the Supporting Information. Assignments of Raman modes are shown in Tables 3 and S2 (Supporting Information). The lower Table 3. Vibrational Frequencies and Assignments of Raman Modes in Neat (MeCN)2−LiTFSI Solvate Electrolyte peak label

Raman shift (cm−1)

species

assignment

a

742

TFSI−

a′

749

TFSI−

b b′

2256 2278

MeCN MeCN

c c′

2294 2308

MeCN MeCN

νs(S−N−S), δ(SO2), δ(CF3) mode a, coordinated to Li+ νs(CN) mode b, coordinated to Li+ νs(C−H) mode c, coordinated to Li+

ref 46−49 25, 45, 47 51, 52 25, 45, 46, 53 51, 52 25

wavenumber region shows vibrational modes related to the TFSI− anion, modes a and a′, and the higher wavenumber region shows vibrational modes related to MeCN, modes b and c. In the case of TFSI−, mode a corresponds to the expansion and contraction of the entire anion and is particularly sensitive to ionic interactions.46−49 It is well-known that the TFSI− anion has two low-energy conformers: a cisoid and a transoid.45,48−50 Due to the structural flexibility of TFSI−, each conformer could coordinate to Li+ in various ways, resulting in multiple overlapping bands around 740−750 cm−1.45 The frequency difference between each anionic coordination structure is only 1−2 cm−1, making it difficult to deconvolute the spectra with high accuracy.45 Therefore, the Raman spectra around 740− 750 cm−1 were fit with two components, a (uncoordinated TFSI−) and a′ (coordinated TFSI−), as an approximation. Mode a is located at ∼742 cm−1 when TFSI− is uncoordinated and is shifted to ∼749 cm−1 as a result of coordination to Li+.25,45−49 The coordinated TFSI− mode is labeled mode a′ in Figure 2. As reported previously, the TFSI− mode in the (MeCN)2−LiTFSI solvate contains contributions from both modes a and a′, with a larger contribution from mode a′, coordinated TFSI−.25,45 Upon addition of the four HFEs to the (MeCN)2−LiTFSI solvate, the ratio of band a to band a′ remains fairly constant, and thus, the coordination environment around TFSI− is unchanged as a result of HFE addition. The same trend is observed for all HFEs tested and agrees well with the previously reported result using TTE as a diluent.25 The coordination of MeCN to the Li+ was studied by interrogating the CN stretching of the MeCN molecule. The high wavenumber region in Figure 2a−d shows the CN stretching vibration, labeled mode b, in addition to the C−H stretching band,51,52 labeled mode c. MeCN coordination to Li+ shifts mode b to higher wavenumbers, resulting in a new mode labeled b′.25,45,46,53 Coordination to Li+ also causes a shift in the C−H stretching mode to mode c′; however, this mode is less sensitive to coordination effects and therefore will not be discussed further. The CN stretching mode is a better proxy to investigate the change in the coordination environment since the coordination to Li+ occurs through the lone pair of N.45 Figure 2 shows the (MeCN)2−LiTFSI solvate complex is mainly composed of coordinated MeCN, mode b′, with a small contribution from free MeCN, mode b. The presence of free MeCN implies that the coordination number of MeCN in the first coordination shell of Li+ is less than 2 or that multimers are

Figure 3. Variable-temperature 7Li T1 curves for the neat (MeCN)2− LiTFSI solvate and the solvate diluted with (a) TTE, (b) OTE, (c) BTFE, and (d) ETE at volume ratios of 2:1, 1:1, and 1:2. The T1 minimum is observed in each sample within the temperature range of −35 to +70 °C. Lines represent the fits of the data to the BPP equation.54

relaxation time, T1, measured as a function of HFE dilution at different temperatures. NMR relaxation measurements are a powerful method to examine the local environment and dynamics of the nucleus at a local level and evaluate the local symmetry around the 7Li nucleus. T1 is the decay constant related to the nuclear spin relaxing back to its equilibrium state in an external magnetic field. 7Li is a quadrupolar nucleus where the charge distribution within the nucleus is nonuniform; 39362

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces therefore, relaxation is affected by the changes in the symmetry of the electron density around Li. Quadrupolar nuclei distribute energy through the quadrupolar relaxation mechanism, and this mechanism is assumed to be the dominating relaxation mechanism.54,55 Therefore, for 7Li, changes in T1 can be regarded as arising from changes in the local bonding environment of Li. T1, however, is sensitive not only to changes in the local symmetry but also to changes in the mobility of the complex. As HFE is added to the solvate, the viscosity of the solution decreases significantly, causing a change in T1. Variable-temperature 7Li T1 experiments were performed to decouple these two effects. A shift of the T1 minimum to lower temperature ranges reflects an increase in mobility, and a shift in the T1 minimum to faster time scales (i.e., T1 values become shorter) indicates a change in the local environment of the nucleus.54,55 Figure 3 shows the variable-temperature T1 curves for the (MeCN)2−LiTFSI solvate electrolyte and (MeCN)2−LiTFSI diluted with HFE cosolvents at volume ratios of 2:1, 1:1, and 1:2. Overall, the T1 minimum of the curve shifts to the lower temperature range with increasing HFE content for all HFEs, indicating enhanced mobility of the complex due to the lower viscosity of the HFE-diluted solutions. Qualitatively, the magnitude of the temperature shift at which the T1 minimum is observed relative to the T1 minimum in the solvate (i.e., shift of the T1 curve along the x-axis) as a result of HFE addition follows the order BTFE ≈ ETE > TTE > OTE. This implies that the order of viscosity of the HFE-added solvates is OTE > TTE > BTFE ≈ ETE, consistent with the measured viscosity value shown in Table 2. The addition of HFEs to the (MeCN)2−LiTFSI also results in a shift of the T1 minimum to faster time scales along the yaxis, suggesting a change in the local coordination environment around Li+. HFE addition to the solvate modifies the Li+ local environment so that the local symmetry around Li becomes more asymmetric, as evidenced by the shift of the T1 minimum to shorter times with increased HFE content. Such a shift could be caused by the replacement of coordinated MeCN by HFE molecules which have high electron density around O and F atoms, resulting in the asymmetry of electron density around Li+. The shift of the T1 minimum along the time axis is observed for all HFEs used, but the shift is more pronounced in the case of TTE- and OTE-added solvates, while relatively small shifts were induced by BTFE and ETE addition as shown in Figure 3. Thus, it is likely that a fraction of HFE coordinates to Li+ likely at the expense of MeCN coordination and that TTE and OTE exhibit a higher tendency to coordinate to Li+ relative to the other two HFEs. For quadrupolar 7Li nuclei experiencing isotropic reorientational diffusion, T1 can be modeled by Bloembergen, Purcell, and Pound (BPP) theory, given in the following equation:54,55 ⎞ ωq 2 ⎛ 4τc τc 1 ⎟ ⎜ = + 2 2 2 2 T1 50 ⎝ 1 + ω0 τc 1 + 4ω0 τc ⎠

τc = τ0 exp(Ea /RT )

(2)

where τ0 is the correlation time constant, Ea is the activation energy for molecular reorientation, and R is the gas constant. According to eq 1, T1 exhibits a minimum at ω0τc = 0.616, which means the correlation time where the minimum occurs is constant.56 The T1 minimum obtained from fitting the data in Figure 3 was used to calculate ωq for all of the electrolytes. τc was calculated at each temperature from the experimental T1 value. The fit parameters give more detailed information about the system. We note that more accurate values can be obtained by fitting data obtained over a broader temperature window. Figure 4 shows the fit parameters of the BPP equation to the experimental T1 curves shown in Figure 3. Figure 4a shows the

Figure 4. Fit parameters of the BPP equation to 7Li variabletemperature T1 measurements. The (a) quadrupolar coupling constant, ωq, and (b) correlation time, τc, at 298 K of the 7Li nuclei in neat (MeCN)2−LiTFSI solvate and the solvate diluted with TTE, OTE, BTFE, and ETE are shown as a function of the HFE content.

quadrupolar coupling constant in units of frequency as a function of the HFE content. In general, for all HFEs examined, ωq increases as more HFE is titrated into the solvate solution (Figure 4a). An increase in ωq suggests larger oscillating electric field gradients (EFGs) around Li+ and therefore efficient exchange of energy between Li+ and its surroundings. We note here that the slope is larger with TTE- and OTE-added solvate solutions than electrolytes with BTFE and ETE cosolvents. The larger ωq in the TTE and OTE systems was also confirmed by shorter T1 values, as large EFGs result in fast relaxation and short T1 processes. These data taken together suggest addition of TTE and OTE induces more asymmetry of electron density around Li+ relative to the (MeCN)2−LiTFSI solvate complex by replacing either the coordinated MeCN or TFSI− anion. Raman spectroscopy suggests that the coordination of the TFSI− anion is relatively unchanged after HFE addition, whereas free MeCN was released as HFE was added to the solvate. Therefore, we hypothesize that the change in local symmetry around Li+ is the result of HFE coordination and the release of free MeCN. Figure 4b shows the correlation time of the (MeCN)2− LiTFSI solvate and its HFE-added solutions at 298 K. The 7Li correlation time for TTE and OTE is again higher relative to that of the other two HFEs. The higher correlation time is associated with worse Li+ ion mobility in the solvate electrolyte in which TTE and OTE were added relative to those in which

(1)

where ωq is the quadrupolar coupling constant, τc is the correlation time, and ω0 is the Larmor frequency. The quadrupolar coupling constant, ωq, is a measure of the energy of interaction between the nuclear quadrupolar moment, which is an intrinsic property of the nucleus, and the electric field gradients (EFGs) generated by its surroundings.55 The correlation time, τc, for molecular reorientation is defined in the following equation:55 39363

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

Figure 5. (a−d) 15N NMR spectra of a 2:1 mixture (by volume) of MeCN:HFE and solvate:HFE, as indicated. 15N chemical shifts were referenced to NH415NO3 in D2O at −4 ppm. (e) The displacements in 15N chemical shifts (Δδ) induced by an addition of Li+ salt with respect to the MeCN:HFE mixtures are shown. The NMR spectra shown here were acquired at room temperature.

Figure 6. Quantum chemical calculations to understand the effect of HFE addition on the coordination environment around Li+ in (MeCN)2− LiTFSI solvate electrolytes. (a) Atomic configuration of the solvation shell around Li+ in neat (MeCN)2−LiTFSI solvate obtained from AIMD simulations. (b−e) Most stable atomic configuration upon replacing one molecule of MeCN shown in (a) with one TTE, OTE, BTFE, and ETE molecule, respectively, obtained from QC calculations. (f) Enthalpy change associated with replacing one MeCN molecule in the Li+ solvation shell of (MeCN)2−LiTFSI solvate with different HFE molecules (ΔHrepl) in a dielectric medium at 298 K.

probe the coordination of MeCN to the Li+ with high sensitivity (Figure 5). Parts a−d of Figure 5 show the 15N NMR spectra of the solvate:HFE electrolyte solutions, with the corresponding MeCN:HFE solutions without LiTFSI salt as a reference in each case. Figure 5a shows the case in which the resonance of the 15N nucleus shifts upfield by 20.535 ppm as a result of LiTFSI addition to the MeCN:TTE solution. The shift is due to the shielding effect of Li+ on the N lone pair as a result of MeCN coordination to Li+.57 The displacement relative to the solution without any salt, in this case the neat MeCN:TTE, is given by Δδ, which can be used as a proxy to indicate the extent of Li+−MeCN coordination.57 We note the large viscosity difference between the neat MeCN and (MeCN)2−LiTFSI solution could also affect the 15N chemical shift (Figure S4, Supporting Information), and this system is not discussed here. Instead, the viscosity difference between MeCN:HFE and

BTFE and ETE were added, consistent with the higher viscosity of TTE and OTE solutions reported in Table 2. 15 N NMR Spectroscopy. 15N NMR was used to further verify the effect of HFE addition on the coordination environment of MeCN in the solvate electrolyte. Although the low natural abundance of 15N (0.365%) poses a challenge to collecting high signal to noise ratio spectra, the 15N spin 1/2 nucleus yields sharp NMR lines, and its wide chemical shift range and cation affinity make it a good probe for ionic coordination.57 Raman spectroscopy and variable-temperature T1 measurements suggest that both the quantity and types of HFE cosolvent affect the equilibrium between free and coordinated MeCN in the solvate electrolyte. Therefore, it is worthwhile to examine the Li+−MeCN interaction directly by monitoring changes in the chemical shift of the 15N nucleus in MeCN as a function of LiTFSI addition and HFE dilution. The 15N chemical shifts of the MeCN solvate and its HFE diluents were used as a proxy to 39364

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

MeCN molecule in the (MeCN)2−LiTFSI solvate structure (obtained from AIMD simulation shown in Figure 6a) with different HFEs. In each configuration, the HFE molecule coordinates with Li+ through the ethereal O atom (with a Li−O separation distance of ∼2.1−2.3 Å). This preferential association of Li+ with ethereal O is a manifestation of the relative partial charge distribution in the HFE molecules. For all four HFEs, our QC calculations show that the ethereal O atoms possess the highest (negative) partial charge in the molecule with values of −0.75 to −0.91 (in comparison, F atoms have an average charge of ∼−0.54). The atomic partial charge in each HFE is shown in Figure S5 (Supporting Information). This negative partial charge on the O atom makes it the most electronegative atom in the HFE and, consequently, the preferred site for Li+ binding. In addition, formation of these MeCN−LiTFSI−HFE complexes is reasonable since their computed solvation free energies ΔGsolv (TTE, − 14.76 kcal mol−1; OTE, − 15.22 kcal mol−1; BTFE, − 14.76 kcal mol−1; ETE, − 9.45 kcal mol−1) are comparable to that of the (MeCN)2−LiTFSI solvate (ΔGsolv ≈ − 16.37 kcal mol−1). Figure 6f provides the enthalpy change associated with replacing an MeCN molecule in the (MeCN)2−LiTFSI solvate structure with each HFE. Evidently, replacing an MeCN molecule with BTFE or ETE is more endothermic as compared to that with TTE and OTE. This makes replacement of MeCN with TTE/OTE more energetically favorable (i.e., more likely) as compared to that with BTFE or ETE. Therefore, diluting the (MeCN)2−LiTFSI solvate with TTE or OTE results in a higher amount of free MeCN as compared to that with BTFE or ETE, consistent with the 15N NMR experiments shown in Figure 5. The formation of free MeCN in solvate electrolytes could affect the battery cycling performance in different ways. In the presence of free MeCN, the kinetics of the S8 reduction/ oxidation processes are improved by the capability of free MeCN to solubilize polysulfide intermediates.25 Instead of forcing the S8 reduction/oxidation pathway to follow a putative solid-state conversion reaction, the presence of a polysulfide solvating molecule, MeCN, facilitates the S8 reaction kinetics.25 The improved kinetics at the electrolyte/electrode interface are likely to result in good battery cyclability. However, the presence of free MeCN could also affect other cell functionalities. For example, MeCN can affect the stability of the Li metal anode through the spontaneous reaction of Li metal with MeCN to form an unstable SEI during long-term battery cycling.58 Additionally, free MeCN in the solvate, while improving the S8 reduction/oxidation kinetics, could also affect polysulfide dissolution and diffusion. Therefore, it is important to examine the effect of free MeCN on the Li metal anode and on the polysulfide solubility to deconvolute the beneficial effect of TTE and OTE cosolvents. Li Metal Anode Stability and Lithium Polysulfide Solubility in the HFE-Diluted MeCN Solvate Electrolytes. Li−Li symmetric cells with solvate and solvate:HFE (2:1) mixtures were tested by galvanostatic cycling to investigate the cycling stability of the Li metal anodes (Figure S6, Supporting Information). The voltage profile of the Li−Li symmetric cell with HFE-free neat solvate exhibits unstable plating/stripping behavior resulting in cell death after a few cycles, possibly due to the high viscosity of the electrolyte (Table 2). High viscosity of the electrolyte hinders facile Li+ ion transport, causing irreversible Li plating/stripping.59,60 Likely, the quick capacity fading of the Li−S cell with (MeCN)2−LiTFSI (Figure 1) can be attributed to the unfavorable Li plating/stripping at the

solvate:HFE is negligible; therefore, any chemical shift displacement is the result of changes in the local bonding environment of N atoms in MeCN. Figure 5e shows that the displacements relative to the neat MeCN:HFE solutions (Δδ) are smaller in TTE and OTE than those observed in BTFE and ETE, which indicates that the degree of Li+−MeCN coordination is less in TTE- and OTEadded solutions. This is the consequence of a stronger interaction between Li+ and TTE/OTE than other HFEs, resulting in a displacement of coordinated MeCN from the solvate structure and a release of free MeCN in solution. In other words, the smaller the Δδ, the higher the free MeCN content, which is in the order OTE > TTE ≫ ETE ≈ BTFE. This suggests that the TTE- and OTE-added solvates have more free MeCN than the solvates with BTFE and ETE. We note that the exchange between MeCN and HFE is fast on the time scale of NMR experiments resulting in a single 15N resonance, which is the concentration-weighted average of free MeCN and coordinated MeCN. TTE and OTE exchange with coordinated MeCN more readily, resulting in higher free MeCN in solution. Ab Initio Molecular Dynamics Simulations and Quantum Chemical Calculations. To further investigate the effect of addition of HFE on the coordination environment of Li+ in (MeCN)2−LiTFSI solvate electrolytes, we performed ab initio molecular dynamics (AIMD) simulations and quantum chemical (QC) cluster calculations. AIMD simulations in the neat (MeCN)2−LiTFSI solvate indicate that the first solvation shell around almost every Li+ is composed of one TFSI− anion and two MeCN molecules (Figure 6a), consistent with previous spectroscopic and theoretical studies.25,45 In this solvation complex, the Li+ ions coordinate with MeCN through the N atom (Li−N distance ∼2.1−2.3 Å), while they coordinate with the TFSI− anion through the O atoms of the sulfonyl group (Li−O separation ∼1.98−2.1 Å). A minor fraction of TFSI− anions bind with more than one Li+ center, which results in a small amount of uncoordinated MeCN, in good agreement with previous investigations.25,45 Consequently, a small fraction of MeCN molecules (∼8%) are found to be uncoordinated in neat (MeCN)2−LiTFSI solvate from the AIMD simulations at 300 K. Upon addition of HFE to (MeCN)2−LiTFSI, we observe that the HFE molecules replace the MeCN in the solvation shell around Li+ without affecting the TFSI coordination around Li+ and, in turn, release free MeCN. This is consistent with the Raman spectroscopy results reported in this work (Figure 2), as well as our recent work on the TTEdiluted (MeCN)2−LiTFSI solvate system.25 The nature of the HFE diluent, however, significantly affects the amount of MeCN released (in this paper, the amount of “released” MeCN is obtained by counting the MeCN molecules whose N atoms are at least 2.7 Å away from each Li+ in the system). AIMD simulations on the (MeCN)2−LiTFSI solvate diluted with HFEs at a 1:1 volume ratio show that BTFE and ETE release ∼43% and ∼28% free MeCN, repectively. On the other hand, higher fractions of free MeCN are released in electrolytes diluted with TTE (∼57%) and OTE (∼50%). This is qualitatively in good agreement with the 15N NMR studies (Figure 5). Next, we identify the thermodynamic origin for the higher tendency of TTE and OTE (as compared to BTFE and ETE) to release free MeCN in the (MeCN)2−LiTFSI solvate using QC calculations. Parts b−e of Figure 6 show the most stable configuration of the Li+ solvation complex upon replacing one 39365

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces

the solution.14,62,63 Additional absorbance over a relatively wide wavelength range is also observed and can be attributed to the presence of various polysulfide species that are expected due to disproportionation reactions and complex equilibria of several polysulfide species in solution.20 The intensity of the absorption band at 420 nm is larger in BTFE- and ETEadded solvates, suggesting higher polysulfide solubility in these solutions. The absorbance is negligible in neat HFE solutions saturated with Li2S8 (Figure S7, Supporting Information), suggesting the solution structure of the solvate electrolyte and its equilibrium with lithium polysulfides affect the polysulfide solubility. Interestingly, the higher free or uncoordinated MeCN in TTE- and OTE-added solutions does not necessarily translate into higher solubility of polysulfides, as the addition of lithium polysulfides to the (MeCN)2−LiTFSI likely changes the equilibrium among Li+, the TFSI− anion, and the solvent molecules. The change in the Li+ solvation structure as a result of lithium polysulfide addition will not be discussed here. To understand the effect of the polysulfide solubility in (MeCN)2−LiTFSI/HFE electrolytes on the electrochemical performance of the Li−S cells, the chemical composition of the SEI layer on Li metal anodes was analyzed by XPS. Figure 8

anode. Figure S6 shows that the HFE-added solvate electrolytes also exhibit unstable voltage hysteresis for the first few cycles, but the Li−Li plating/stripping behavior becomes more stable upon extended cycling. The TTE and BTFE cells, however, exhibit similar Li−Li plating/stripping behavior, but very different Li−S cycling behavior. Figure 1 shows that the discharge capacity of the Li−S cell incorporating BTFE faded much more quickly than that of the cell incorporating TTE. Therefore, the effect of the electrolyte on the cycling performance must be more complicated than the effect of the electrolyte on the stability of the anode. The presence of higher free MeCN in TTE- and OTE-diluted solvate solutions might imply poor stability of the Li anode affecting the Li plating/ stripping efficiency. The Li plating/stripping behavior in TTE and OTE solvents, however, was similar to or slightly better than that observed in the other two cosolvents. In fact, HFE competes with MeCN for Li+ coordination, and the exchange between HFE and MeCN is in dynamic equilibrium.25 Consequently, this “free” MeCN may not react with Li metal directly. In addition, as suggested in previous work, TFSI− anions are preferentially decomposed on Li metal to form a TFSI-derived SEI layer.22,61 This SEI is believed to inhibit the decomposition of free MeCN.22,61 Therefore, the amount of free MeCN alone does not determine anode stability. Another possible origin of variations in the Li−S cell cycling data as different HFEs are added to the electrolyte is changes in the lithium polysulfide solubility. Previous work using (MeCN)2−LiTFSI/TTE electrolyte showed that delicate control of the polysulfide solubility by temperature affects the redox kinetics and therefore changes the S8 reaction pathway.27 The use of different HFEs could also dictate the polysulfide solubility, resulting in different cycling capabilities. Therefore, evaluating the polysulfide solubility in different HFE systems is important as a major function of the solvate electrolyte with HFE cosolvent is to control this solubility in the Li−S battery. As reported previously, the solubility of lithium polysulfides in the TTE solvent is low, making it difficult to measure using titration.21 Alternatively, the electrolyte solutions saturated with solid “Li2S8” prepared by evaporating the solvent from a solution of Li2S and S8 at the corresponding ratio may be analyzed qualitatively using UV−vis spectroscopy. In addition, the room temperature solubility of polysulfides is too low to probe with UV−vis spectroscopy.27 Therefore, the Li2S8saturated solution was prepared at 70 °C to promote the dissolution. Figure 7a shows a photograph of solvate:HFE (2:1) solutions saturated with Li2S8, and Figure 7b shows the corresponding the UV−vis spectra of each solution. In Figure 7b, the absorbance at ∼420 nm is due to the presence of S42− in

Figure 8. XPS analysis of the SEI layer on Li metal anodes formed after a 100 cycle Li−S galvanostatic cycling experiment in (a) solvate:TTE (2:1), (b) solvate:OTE (2:1), (c) solvate:BTFE (2:1), and (d) solvate:ETE (2:1) electrolytes. The S 2p XPS spectra are shown, including peak fitting and assignments: TFSI (R−SO2−R′) (red), SO32− (blue), S2O62− (purple), bridging sulfur SB0 (green), and terminal sulfur ST−1 (orange).

shows S 2p XPS spectra obtained from Li metal anodes cycled in solvate:HFE (2:1) electrolyte. S 2p XPS exhibits a 2p3/2 and 2p1/2 doublet. The spectra in Figure 8 are fit with this doublet maintaining the 2:1 area ratio, the 1.18 eV energy difference, and equal fwhm between the doublet pairs. Only the binding energy of the high-intensity 2p3/2 component will be described hereafter. In Figure 8b−d, the S 2p region is best fit with four different S binding energies, while in Figure 8a, three different binding energies are found. The two S 2p3/2 peaks at 169.0 and 167.1 eV represent products of LiTFSI salt decomposition, corresponding to the sulfone group (R−SO2−R′) of the TFSI− anion and sulfite (SO32−), respectively.64−66 These R−SO2−R′

Figure 7. (a) Photograph showing solvate:HFE (2:1) solutions saturated with “Li2S8” and (b) the corresponding UV−vis spectra. 39366

DOI: 10.1021/acsami.7b11566 ACS Appl. Mater. Interfaces 2017, 9, 39357−39370

Research Article

ACS Applied Materials & Interfaces and SO32− species are present in all HFEs tested. The additional peak at 165.8 eV representing S2O62− is observed in Li anodes cycled in solvate:TTE electrolyte.67,68 The additional S 2p3/2 peaks in the lower binding energy regime (